Suppression of gene expression by double-stranded RNA (dsRNA) has been demonstrated in a variety of systems including plants (post-transcriptional gene suppression) (Napoli et al., 1990), fungi (quelling) (Romano and Marcino, 1992), and nematodes (RNA interference) (Fire et al., 1998). Double-stranded RNA (dsRNA) is significantly more stable than single-stranded RNA (ssRNA). This difference is pronounced in the intracellular environment (Raemdonck et al., 2006). However, unmodified siRNAs are rapidly degraded in serum, which is a fairly nuclease rich environment. Chemical modification can significantly stabilize the siRNA and improve potency both in vitro and in vivo. Extensive medicinal chemistry has been done over the past 20 years for applications where synthetic nucleic acids are used for experimental or therapeutic applications in vivo, such as in the antisense and ribozyme fields, and hundreds of compounds have been tested in a search for modifications that improve nuclease stability, increase binding affinity, and sometimes also improve pharmacodynamic properties of synthetic nucleic acids (Matteucci, 1997; Manoharan, 2002; Kurreck, 2003). Many of these modifications have already been tested and found to have utility as modifiers for use in traditional 21 mer siRNAs. Several reviews have provided summaries of recent experience with 21 mer siRNAs and chemical modifications (Zhang et al., 2006; Nawrot and Sipa, 2006; Rana, 2007). Modification patterns have not been tested or optimized for use in longer RNAs, such as Dicer-substrate siRNAs (DsiRNAs).
Early attempts to suppress gene expression using long dsRNAs in mammalian systems failed due to activation of interferon pathways that do not exist in lower organisms. Interferon responses are triggered by dsRNAs (Stark et al., 1998). In particular, the protein kinase PKR is activated by dsRNAs of greater than 30 bp long (Manche et al., 1992) and results in phosphorylation of translation initiation factor eIF2α which leads to arrest of protein synthesis and activation of 2′5′-oligoadenylate synthetase (2′-5′-OAS), which leads to RNA degradation (Minks et al., 1979).
In Drosophila cells and cell extracts, dsRNAs of 150 bp length or greater were seen to induce RNA interference while shorter dsRNAs were ineffective (Tuschl et al., 1999). Long double-stranded RNA, however, is not the active effecter molecule; long dsRNAs are degraded by an RNase III class enzyme called Dicer (Bernstein et al., 2001) into very short 21-23 bp duplexes that have 2-base 3′-overhangs (Zamore et al., 2000). These short RNA duplexes, called siRNAs, direct the RNAi response in vivo and transfection of short chemically synthesized siRNA duplexes of this design permits use of RNAi methods to suppress gene expression in mammalian cells without triggering unwanted interferon responses (Elbashir et al., 2001a). The antisense strand of the siRNA duplex serves as a sequence-specific guide that directs activity of an endoribonuclease function in the RNA induced silencing complex (RISC) to degrade target mRNA (Martinez et al., 2002).
In studying the size limits for RNAi in Drosophila embryo extracts in vitro, a lower threshold of around 38 bp double-stranded RNA was established for activation of RNA interference using exogenously supplied double-stranded RNA and duplexes of 36, 30, and 29 bp length (Elbashir et al., 2001b). The short 30-base RNAs were not cleaved into active 21-23-base siRNAs and therefore were deemed inactive for use in RNAi (Elbashir et al., 2001b). Continuing to work in the Drosophila embryo extract system, the same group later carefully mapped the structural features needed for short chemically synthesized RNA duplexes to function as siRNAs in RNAi pathways. RNA duplexes of 21-bp length with 2-base 3′-overhangs were most effective, duplexes of 20, 22, and 23-bp length had slightly decreased potency but did result in RNAi mediated mRNA degradation, and 24 and 25-bp duplexes were inactive (Elbashir et al., 2001c).
Some of the conclusions of these earlier studies may be specific to the Drosophila system employed. Other investigators established that longer siRNAs can work in human cells. However, duplexes in the 21-23-bp range have been shown to be more active and have become the accepted design (Caplen et al., 2001). Essentially, chemically synthesized duplex RNAs that mimicked the natural products that result from Dicer degradation of long duplex RNAs were identified to be the preferred compound for use in RNAi. Approaching this problem from the opposite direction, investigators studying size limits for RNAi in Caenorhabditis elegans found that although a microinjected 26-bp RNA duplex could function to suppress gene expression, it required a 250-fold increase in concentration compared with an 81-bp duplex (Parrish et al., 2000).
Despite the attention given to RNAi research recently, the field is still in the early stages of development. Not all siRNA molecules are capable of targeting the destruction of their complementary RNAs in a cell. As a result, complex sets of rules have been developed for designing RNAi molecules that will be effective. Those having skill in the art expect to test multiple siRNA molecules to find functional compositions. (Ji et al., 2003) Some artisans pool several siRNA preparations together to increase the chance of obtaining silencing in a single study. (Ji et al., 2003) Such pools typically contain 20 nM of a mixture of siRNA oligonucleotide duplexes or more (Ji et al., 2003), despite the fact that a siRNA molecule can work at concentrations of 1 nM or less (Holen et al., 2002). This technique can lead to artifacts caused by interactions of the siRNA sequences with other cellular RNAs (“off target effects”). (Scherer et al., 2003) Off target effects can occur when the RNAi oligonucleotides have homology to unintended targets or when the RISC complex incorporates the unintended strand from and RNAi duplex. (Scherer et al., 2003) Generally, these effects tend to be more pronounced when higher concentrations of RNAi duplexes are used. (Scherer et al., 2003)
In addition, the duration of the effect of an effective RNAi treatment is limited to about 4 days (Holen et al., 2002). Thus, researchers must carry out siRNA experiments within 2-3 days of transfection with an siRNA duplex or work with plasmid or viral expression vectors to obtain longer term silencing.
One major factor that inhibits the effect of siRNAs is the degradation of siRNAs by nucleases. A 3′-exonuclease is the primary nuclease activity present in serum and modification of the 3′-ends of antisense DNA oligonucleotides is crucial to prevent degradation (Eder et al., 1991). An RNase-T family nuclease has been identified called ERI-1 which has 3′→5′ exonuclease activity that is involved in regulation and degradation of siRNAs (Kennedy et al., 2004; Hong et al., 2005). This gene is also known as Thex1 (NM—02067) in mice or THEX1 (NM—153332) in humans and is involved in degradation of histone mRNA; it also mediates degradation of 3′-overhangs in siRNAs, but does not degrade duplex RNA (Yang et al., 2006). It is therefore reasonable to expect that 3′-end-stabilization of siRNAs will improve stability.
XRN1 (NM—019001) is a 5′→3′ exonuclease that resides in P-bodies and has been implicated in degradation of mRNA targeted by miRNA (Rehwinkel et al., 2005) and may also be responsible for completing degradation initiated by internal cleavage as directed by a siRNA. XRN2 (NM—012255) is a distinct 5′→3′ exonuclease that is involved in nuclear RNA processing. Although not currently implicated in degradation or processing of siRNAs and miRNAs, these both are known nucleases that can degrade RNAs and may also be important to consider.
RNase A is a major endonuclease activity in mammals that degrades RNAs. It is specific for ssRNA and cleaves at the 3′-end of pyrimidine bases. SiRNA degradation products consistent with RNase A cleavage can be detected by mass spectrometry after incubation in serum (Turner et al., 2007). The 3′-overhangs enhance the susceptibility of siRNAs to RNase degradation. Depletion of RNase A from serum reduces degradation of siRNAs; this degradation does show some sequence preference and is worse for sequences having poly A/U sequence on the ends (Haupenthal et al., 2006). This suggests the possibility that lower stability regions of the duplex may “breathe” and offer transient single-stranded species available for degradation by RNase A. RNase A inhibitors can be added to serum and improve siRNA longevity and potency (Haupenthal et al., 2007).
In 21 mers, phosphorothioate or boranophosphate modifications directly stabilize the internucleoside phosphate linkage. Boranophosphate modified RNAs are highly nuclease resistant, potent as silencing agents, and are relatively non-toxic. Boranophosphate modified RNAs cannot be manufactured using standard chemical synthesis methods and instead are made by in vitro transcription (IVT) (Hall et al., 2004 and Hall et al., 2006). Phosphorothioate (PS) modifications can be easily placed in the RNA duplex at any desired position and can be made using standard chemical synthesis methods. The PS modification shows dose-dependent toxicity, so most investigators have recommended limited incorporation in siRNAs, favoring the 3′-ends where protection from nucleases is most important (Harborth et al., 2003; Chiu and Rana, 2003; Braasch et al., 2003; Amarzguioui et al., 2003). More extensive PS modification can be compatible with potent RNAi activity; however, use of sugar modifications (such as 2′-O-methyl RNA) may be superior (Choung et al., 2006).
A variety of substitutions can be placed at the 2′-position of the ribose which generally increases duplex stability (Tm) and can greatly improve nuclease resistance. 2′-O-methyl RNA is a naturally occurring modification found in mammalian ribosomal RNAs and transfer RNAs. 2′-O-methyl modification in siRNAs is known, but the precise position of modified bases within the duplex is important to retain potency and complete substitution of 2′-O-methyl RNA for RNA will inactivate the siRNA. For example, a pattern that employs alternating 2′-O-methyl bases can have potency equivalent to unmodified RNA and is quite stable in serum (Choung et al., 2006; Czauderna et al., 2003).
The 2′-fluoro (2′-F) modification is also compatible with siRNA function; it is most commonly placed at pyrimidine sites (due to reagent cost and availability) and can be combined with 2′-O-methyl modification at purine positions; 2′-F purines are available and can also be used. Heavily modified duplexes of this kind can be potent triggers of RNAi in vitro (Allerson et al., 2005; Prakash et al., 2005; Kraynack and Baker, 2006) and can improve performance and extend duration of action when used in vivo (Morrissey et al., 2005a; Morrissey et al., 2005b). A highly potent, nuclease stable, blunt 19 mer duplex containing alternative 2′-F and 2′-O-Me bases is taught by Allerson. In this design, alternating 2′-O-Me residues are positioned in an identical pattern to that employed by Czauderna, however the remaining RNA residues are converted to 2′-F modified bases. A highly potent, nuclease resistant siRNA employed by Morrissey employed a highly potent, nuclease resistant siRNA in vivo. In addition to 2′-O-Me RNA and 2′-F RNA, this duplex includes DNA, RNA, inverted abasic residues, and a 3′-terminal PS internucleoside linkage. While extensive modification has certain benefits, more limited modification of the duplex can also improve in vivo performance and is both simpler and less costly to manufacture. Soutschek et al. (2004) employed a duplex in vivo and was mostly RNA with two 2′-O-Me RNA bases and limited 3′-terminal PS internucleoside linkages.
Locked nucleic acids (LNAs) are a different class of 2′-modification that can be used to stabilize siRNAs. Patterns of LNA incorporation that retain potency are more restricted than 2′-O-methyl or 2′-F bases, so limited modification is preferred (Braasch et al., 2003; Grunweller et al., 2003; Elmen et al., 2005). Even with limited incorporation, the use of LNA modifications can improve siRNA performance in vivo and may also alter or improve off target effect profiles (Mook et al., 2007).
Synthetic nucleic acids introduced into cells or live animals can be recognized as “foreign” and trigger an immune response. Immune stimulation constitutes a major class of off-target effects which can dramatically change experimental results and even lead to cell death. The innate immune system includes a collection of receptor molecules that specifically interact with DNA and RNA that mediate these responses, some of which are located in the cytoplasm and some of which reside in endosomes (Marques and Williams, 2005; Schlee et al., 2006). Delivery of siRNAs by cationic lipids or liposomes exposes the siRNA to both cytoplasmic and endosomal compartments, maximizing the risk for triggering a type 1 interferon (IFN) response both in vitro and in vivo (Morrissey et al., 2005b; Sioud and Sorensen, 2003; Sioud, 2005; Ma et al., 2005). RNAs transcribed within the cell are less immunogenic (Robbins et al., 2006) and synthetic RNAs that are immunogenic when delivered using lipid-based methods can evade immune stimulation when introduced unto cells by mechanical means, even in vivo (Heidel et al., 2004). However, lipid based delivery methods are convenient, effective, and widely used. Some general strategy to prevent immune responses is needed, especially for in vivo application where all cell types are present and the risk of generating an immune response is highest. Use of chemically modified RNAs may solve most or even all of these problems.
Although certain sequence motifs are clearly more immunogenic than others, it appears that the receptors of the innate immune system in general distinguish the presence or absence of certain base modifications which are more commonly found in mammalian RNAs than in prokaryotic RNAs. For example, pseudouridine, N6-methyl-A, and 2′-O-methyl modified bases are recognized as “self” and inclusion of these residues in a synthetic RNA can help evade immune detection (Kariko et al., 2005). Extensive 2′-modification of a sequence that is strongly immunostimulatory as unmodified RNA can block an immune response when administered to mice intravenously (Morrissey et al., 2005b). However, extensive modification is not needed to escape immune detection and substitution of as few as two 2′-O-methyl bases in a single strand of a siRNA duplex can be sufficient to block a type 1 IFN response both in vitro and in vivo; modified U and G bases are most effective (Judge et al., 2006). As an added benefit, selective incorporation of 2′-O-methyl bases can reduce the magnitude of off-target effects (Jackson et al., 2006). Use of 2′-O-methyl bases should therefore be considered for all siRNAs intended for in vivo applications as a means of blocking immune responses and has the added benefit of improving nuclease stability and reducing the likelihood of off-target effects.
Although cell death can results from immune stimulation, assessing cell viability is not an adequate method to monitor induction of IFN responses. IFN responses can be present without cell death, and cell death can result from target knockdown in the absence of IFN triggering (for example, if the targeted gene is essential for cell viability). Relevant cytokines can be directly measured in culture medium and a variety of commercial kits exist which make performing such assays routine. While a large number of different immune effector molecules can be measured, testing levels of IFN-α, TNF-α, and IL-6 at 4 and 24 hours post transfection is usually sufficient for screening purposes. It is important to include a “transfection reagent only control” as cationic lipids can trigger immune responses in certain cells in the absence of any nucleic acid cargo. Including controls for IFN pathway induction should be considered for cell culture work. It is essential to test for immune stimulation whenever administering nucleic acids in vivo, where the risk of triggering IFN responses is highest.
There is therefore a need to provide a chemical modification pattern that would enhance the efficacy of a dsRNA, particularly DsiRNA. The modifications 1) should not reduce potency; 2) should not interfere with Dicer processing; 3) should improve stability in biological fluids (reduce nuclease sensitivity); 4) should block or evade detection by the innate immune system; 5) should not be toxic; and 6) should not increase cost or impact ease of manufacturing.
The invention provides compositions useful in RNAi for inhibiting gene expression and provides methods for their use. In addition, the invention provides RNAi compositions and methods designed to maximize potency, enhance Dicer processing, improve stability while evading the immune system and are not toxic. Additionally, various embodiments of the invention are suited for high throughput, small scale synthesis to meet research needs as well as large scale manufacturing for therapeutic applications. These and other advantages of the invention, as well as additional inventive features, will be apparent from the description of the invention provided herein.